Sensitivity-based, non-parametric design optimization under uncertainty for arbitrary objective functions in industrial application

Deterministic sensitivity-based optimization approaches tend to provide designs, which are sensitive to scattering properties such as material parameters, geometric deviations or load direction. Robust design optimization (RDO) techniques provide designs being lightweight and at the same time robust with respect to uncertainties. These uncertainties arise from manufacturing tolerances and material variations as well as boundary condition discrepancies. The designs obtained with RDO may differ significantly from the ones obtained by classical, deterministic optimization approaches. Sensitivity-based RDO approaches are very fast, but require new implementations for each combination of objective function, design variables and random parameters. However, the current talk presents an approach that uses the existing capabilities of Abaqus and Tosca. It is implemented as a plug-on and can easily be applied to various objective functions (like compliance, stress or buckling load) and optimization approaches (like sizing and topology optimization). By application to industrial examples, the capabilities and limitations of the approach are demonstrated.

Options

Sensitivity-based, non-parametric design optimization under uncertainty for arbitrary objective functions in industrial application